1. Field of the Invention
The invention relates to the field of metrology, and in particular, to a system and method for accurately and efficiently measuring semiconductor structure characteristics.
2. Related Art
Integrated circuit (IC) device geometries continue to shrink with each generation of process technology. Those size reductions provide significant efficiency and performance benefits in modern ICs. At the same time, the production of such ICs requires increasingly sophisticated and precise manufacturing processes, which in turn requires that highly sensitive metrology techniques and tools be used to ensure proper manufacturing results. Currently, one of the most effective metrology techniques for modern ICs is scatterometry, in which light scattered from a pattern on a wafer is used to determine physical characteristics for structures formed on that wafer. The two main methods of performing scatterometry are spectroscopic scatterometry and single wavelength scatterometry.
FIG. 1A shows a spectroscopic scatterometry system 100A for performing metrology on a test sample 190A. Spectroscopic scatterometry system 100A includes a broadband light source 110A, focusing optics 120A, a stage 130A for supporting test sample 190A, an order blocking aperture 135A, analyzing optics 140A, a spectrometer 145A, and an array detector 150A. To perform a scatterometry measurement, light source 110A generates a broadband probe beam 111A (i.e., a light beam that includes a wide range of wavelengths, such as a white light), which focusing optics 120A directs onto a measurement pattern 191A on test sample 190A. Note that in some spectroscopic scatterometry systems, focusing optics 120A (and analyzing optics 140A) can include polarizing elements to enable ellipsometric analyses to be performed on the scattered light. Note further that, unlike most optical metrology tools, scatterometry tools such as system 100A generally require a physical pattern on the test sample being measured, so that sufficient light scattering (which is generally due to diffraction effects in the test pattern) occurs.
The scattered light 112A created by diffraction at test sample 190A is composed of a plurality of output beam components called orders, each with its own direction of propagation. Each of the diffraction orders has its own polar angle (angle with respect to wafer normal) and its own azimuthal angle (angle with respect to the projection of the probe beam 111A in the plane of the wafer). There is usually a component emitted with a polar angle equal to the polar angle (angle of incidence) of the probe beam 111A and with an azimuthal angle relative to probe beam 111A of 180 degrees. This component (identified in FIG. 1A as zeroth order beam 112A(0)) is defined as the zeroth diffraction order and corresponds to the reflected beam from an unpatterned sample. The presence of measurement pattern 191A creates other diffraction orders (e.g., negative first order beam 112A(−1) and first order beam 112A(+1)) with other polar angles, both greater than and smaller than the zeroth polar angle. If pattern 191A is one-dimensional and if the projection of the probe beam 111A onto the wafer is aligned with a symmetry axis of pattern 191A, all diffraction orders will have an azimuthal angle of 180 degrees relative to probe beam 111A. Otherwise, orders with other azimuthal angles may also be present. This asymmetric situation is called conical diffraction. In general, the polar (and in some cases azimuthal) angles for all diffraction orders other than the zeroth order depend on wavelength. Therefore all diffraction orders other than the zeroth order are not narrow beams, but instead include a variety of components having various polar and azimuthal angles.
A portion of the scattered light 112A generated in response to broadband light beam 111A is collected by analyzing optics 140A. Often only the zeroth order light is collected (e.g., beam 112A(0), but other orders may also be collected. Zeroth order beam 112A(0) is selected by order blocking aperture 135A, and then passes through analyzing optics 140A, after which it is dispersed by spectrometer 145A onto array detector 150A. Array detector 150A measures a broadband spectrum of intensities for the various wavelengths of light making up broadband light beam 111A. This output spectrum provides a “pattern signature” that is representative of the particular dimensional characteristics of measurement pattern 191A (e.g., dimensions, composition, and surface roughness). By analyzing the broadband spectrum (or spectra) measured by array detector 150A in conjunction with mathematical modeling of measurement pattern 191A, the desired physical characteristic information of measurement pattern 191A can be determined, even if those physical dimensions are smaller than the wavelengths of light in broadband light beam 111A.
FIG. 1B shows a single wavelength scatterometry system 100B for performing metrology on a test sample 190B. Spectroscopic scatterometry system 100B includes a narrowband light source 110B, focusing optics 120B, a stage 130B for supporting test sample 190B, analyzing optics 140B, and an array detector 150B. To perform a scatterometry measurement, light source 110B generates a narrowband light beam 111B (i.e., a light beam that includes a single wavelength, such as a laser light), which focusing optics 120B directs onto a measurement pattern 191B on test sample 190B. As described with respect to scatterometry system 100A shown in FIG. 1A, focusing optics 120B (and analyzing optics 140B) can include polarizing elements to enable ellipsometric analyses to be performed on the scattered light 112B that is scattered from measurement pattern 191B in response to light beam 111B.
The scattered light 112B is directed by analyzing optics 140B onto array detector 150B, which measures the intensity and directions of the light scattering from measurement pattern 191B. In this case the various diffraction orders making up scattered light 112B (e.g., first order beam 112B(+1), zeroth order beam 112B(0), and negative first order beam 112B(−1)) are all narrowband beams and each has a unique polar and azimuthal angle because there is only a narrow range of wavelengths present in narrowband probe beam 111B. Array sensor 150B measures the intensity and positions of some fraction of the diffraction orders. Knowing the position on the array sensor, combined with the properties of analyzing optics 140B, it is possible to extract the polar and azimuthal angle of each detected diffraction order. The arrangement of diffraction orders, their individual intensities, their polarization properties, and their polar and azimuthal angles depend on the properties of pattern 191B and therefore constitute a “pattern signature” for pattern 191B. This pattern signature can be used with mathematical modeling to extract dimensional and other parameters of pattern 191B.
It is possible to extract even more information about the measurement pattern with either the spectroscopic or narrow band systems by measuring with probe beams of different angles. The angles and spectra of the various diffraction orders depend on both the polar angle and azimuthal angle of the probe beam. For instance, it is possible to measure at different probe azimuthal angles by rotating the wafer in its own plane by means of a rotational mechanism incorporated into stage 190A (shown in FIG. 1A) or 190B (shown in FIG. 1B). Measurements can be taken at two or more angles in sequence, rotating the wafer to the desired azimuthal angle before each measurement. In this case the rest of the measurement system can remain stationary. Measurements can also be made at multiple polar angles, but this requires moving at least one or more of the modules of the measurement system.
In all these scatterometry systems, the scattered light has the same wavelength as the probe light. In the spectroscopic systems each wavelength component of the scattered light is created by exactly the same wavelength component of the probe light. In the narrow-band system all of the scattered light is in the same narrow wavelength range of the probe light. The equality of scattered and probe wavelengths is called elastic scattering, due to the fact that the scattered photons have the same energy as the probe photons and no energy is gained or lost to the sample.
Thus, scatterometry (both spectroscopic scatterometry and single wavelength scatterometry) provides metrology capabilities that typically exceed the capabilities of most other non-destructive measurement techniques, and accordingly is the technique of choice for measuring the extremely small semiconductor structures in advanced ICs. However, as scaling of semiconductor devices extends further into the sub-micron range, material properties (i.e., material characteristics other than dimension) such as stress, strain, embedded charge, composition, and degree of crystallinity become increasingly important.
For example, material stress plays a significant role in the performance of the miniature transistors used in advanced ICs. Because material stress is affected by structure size, the stress within, for example, the active region of a MOS transistor cannot be determined from a stress measurement on a bulk region of a wafer. Unfortunately, conventional stress measurement techniques are mainly directed toward bulk measurements (e.g., the measurement of stress within a film formed over an entire wafer), and are therefore not effective for device-level measurements. For example, Raman spectroscopy is one conventional stress measurement technique for measuring stress in silicon (Si) and silicon germanium (SiGe) structures on semiconductor wafers. Raman scattered light usually is composed of several discrete narrow wavelength output beam components, shifted both above and below the incident narrow wavelength range. The magnitude of the wavelength shift of the highest intensity shifted component is determined by the stress level in the silicon. In a silicon germanium structure the shift of this highest intensity shifted component is determined by both the stress within the silicon germanium and the particular silicon germanium composition (other output beam components exhibit intensities and wavelength shifts that are mainly affected by silicon germanium composition). Therefore, the measured shifts and intensities of the various output beam components can be used to determine the stress and composition of Si and SiGe. A similar process can be used to measure other crystalline and polycrystalline materials. Raman spectroscopy has been combined with high resolution microscopy to make measurements with a spatial resolution down to about 0.5 um. However, this level of spatial resolution is not sufficient for making measurements on advanced semiconductor structures that have dimensions of less than 500 nm.
Accordingly, it is desirable to provide a method and system for measuring material properties in miniature devices and structures.